Emeli Dral

Co-founder and CTO @EvidentlyAI, #MachineLearning instructor. Gymnastics, basketball and coffee.

[Evidently AI @EvidentlyAI]

Evidently has 5000 ⭐️ on GitHub 🎉🎉🎉

Thank you for using Evidently and helping us make it a trusted open-source ML and LLM observability framework!

Let AI observability be available to all!

To support the project, give us a ⭐️:
https://t.co/AL9o0YpBcr

[Alexey Grigorev @Al_Grigor]

LLM Zoomcamp has started!

Join us live or watch later in replay:

https://t.co/MpdVpEgVBg

[Emeli Dral @EmeliDral]

🚀 New release: Evaluate LLMs with Evidently open-source!

Reports, Test Suites, and a Monitoring dashboard with lots of methods:
⚖️ LLM-as-a-judge
🤗 ML-based scoring
📐 Semantic similarity
💬 Text stats/patterns

Tutorial: https://t.co/rSMJ1oYBqf
GitHub: https://t.co/i0IUdpkDvR

[Alexey Grigorev @Al_Grigor]

MLOps Zoomcamp starts now!

Join us live or watch later in replay

https://t.co/WaPKMTO8M7

[Emeli Dral @EmeliDral]

And how did you spend your Saturday? 👀

A spoiler preview of mine👇🏽

Working on the new ML Observability Course for @EvidentlyAI!

Register here: https://t.co/2dmuym67DF

[Evidently AI @EvidentlyAI]

🚀 Free Open-source ML observability course!

👩‍🎓 For data scientists, ML engineers, technical PMs, or analysts
⭐️ Open-source focused
🗓 Starts Oct 16
✅ Free

Enroll here: https://t.co/QUQZIi7CQR
Welcome video: https://t.co/vEfrbjv3sk

Details 👇

1/3 🧵

[Evidently AI @EvidentlyAI]

Our CTO, @EmeliDral, will be speaking at the AI Superstream on Sept 13!

Stop by to learn how to evaluate quality and drift in text and multimodal data 🔡

Thanks @OReillyMedia for inviting us and @fab_clemente for hosting the session!

Join for free 👉 https://t.co/ImjAGvKBD9

[Emeli Dral @EmeliDral]

⏰ Staying Alert: how to continuously test ML models in prod.

It was really fun to speak at @PyConDE. They now published the recording of the talks!

🚦 How to automate ML test thresholding?
💻 How to compute complex metrics?
🏗 How should you retrain?

https://t.co/HRgvafu8SW

[Evidently AI @EvidentlyAI]

🛠 How do top engineering teams apply ML to improve products and processes?

We put together a list of 200 case studies that share practical ML use cases and learnings from designing ML systems: https://t.co/Ipe0L1OfB8

Here are our five favorites 👇

1/7 🧵

[Emeli Dral @EmeliDral]

📊Meet Evidently 0.3.2.1!

Evidently HTML Reports finally got more lightweight!

Great news for anyone working with larger datasets!

By default, we will show aggregated plots. If you want the old look, choose a "raw data" render option.

Release: https://t.co/OnxMk9NTnn

[Emeli Dral @EmeliDral]

🚀Meet @EvidentlyAI 0.3.1.

💬 Text descriptors are supported in all relevant metrics! You can run tests like "check that text length is within these bounds."
🔢 Embeddings drift detection in Tests Suites
🐞 A lot of fixes (thanks to everyone reporting!)

https://t.co/SxZBJaLPQj

[Cheuk Ting Ho @cheukting_ho]

@EmeliDral is talking about continuous testing for machine learning models at @PyConDE #PyConDE

[Emeli Dral @EmeliDral]

New release! @EvidentlyAI 0.3 is here!

🆕 Embeddings drift detection with 4 methods
✅ Customize JSON output to define what to include
📚 Use text descriptors in tabular checks (e.g, “test that text length is between X and Y”)
⛔️ Old API removed

Release:
https://t.co/2QpwDJ1m3c

[Evidently AI @EvidentlyAI]

🤔 Why we do ML model retraining wrong, and how to do better

Our CTO, @EmeliDral, spoke on @PyDataGlobal 2022.

She discussed how to design a pragmatic ML model retraining strategy, considering available resources, actions and costs of error👇
https://t.co/W6Yw7dUpiO

[Emeli Dral @EmeliDral]

I’m extremely happy to be part of the event! People here are incredible❤️

[Sam Rhynas @lilacpatsy]

And we’re off! First #Pydata #Edinburgh in three years! Welcome back everyone, and big thanks to wood Mackenzie and @CathcartRecruit

[Emeli Dral @EmeliDral]

The latest release contains a lot more!

📊A new Text Overview preset
📚All relevant Evidently tests and metrics support texts.
🔡Support for multi-model datasets.

Check it out!

GitHub: https://t.co/i0IUdplblp

Release blog: https://t.co/5KDXMRuku3

[Emeli Dral @EmeliDral]

📚How can you evaluate data drift on text data?

We’ve just released the latest functionality in the @EvidentlyAI open-source library that does just that.

Here are the two approaches that we implemented. 👇🏼

(1/4)

[Emeli Dral @EmeliDral]

Text descriptors drift.

You can also look at the shift in the text properties - such as length, out-of-vocabulary words, or non-letter symbols.

You can then compare the distribution of the text descriptors between the two datasets.

(3/4)

[Emeli Dral @EmeliDral]

Meet @Evidently 0.2, a stable release of our open-source ML monitoring tool:

​​👩🏽‍🔧 A major refactoring
🧱 100+ ML tests and metrics
🚦 Simplified API
🛠 Integrations with Airflow and Mlflow

Getting Started tutorial 👇

Video: https://t.co/i0ohJOfS62